Reinventing the Network for AI Workloads
Step into the future of data centers specifically engineered for AI workloads with AMD Pensando™ open networking solutions, a groundbreaking suite of performant, programmable, and open products that enhance and elevate the network from end-to-end.
AMD Pensando™ DPUs Power the Front-End Network
Accelerate data movement and infrastructure services at wire speed, freeing CPU resources for AI workloads.
Scale-Up Networking for Evolving AI Workloads
Modern AI models have outgrown the single GPU. Scale-up networking connects multiple GPUs within a system or rack to form one compute domain, providing the ultra-high bandwidth and low latency that large-scale workloads like agentic AI demands.
Scale-Out and Scale-Across with AI NICs
When AI workloads span servers, racks, and data centers, seamless GPU-to-GPU connectivity becomes critical. AMD AI NIC™ technology powers the high-speed synchronization large-scale AI requires.
AMD Helios: Leadership Rack Performance for Hyperscale AI1
The AMD Helios Rackscale solution design is a fully integrated AI infrastructure, combining the latest AMD Instinct™ GPUs, AMD EPYC™ Server CPUs, and AMD Pensando™ networking, designed using open industry standards enabling large-scale inference, frontier model training and fine-tuning.
AMD Pensando™ Vulcano 800 AI NIC
Complete AI jobs up to 13% faster2 with the only AI NIC on the market to support up to 2.4 Tbps of scale-out bandwidth per GPU.
Enabling AI-Ready Scale-Out Networking for AI Workloads
Learn how scale-out networking improves GPU utilization, accelerates AI jobs, and reduces the cost of scaling AI infrastructure.
Adapters & CPUs
The World Trades on AMD Solarflare™ Low Latency Network Adapters
Purpose-designed silicon, software, and adapters for capital markets
9 out of 10 Global Stock Exchanges use Solarflare Solutions for Trading3
AMD x86 Embedded Processors
AMD x86 embedded processors offer high-throughput compute, robust I/O, and scalability for applications like firewalls, routers, and network appliances – delivering performance, efficiency, and long-term reliability at the edge or in the data center.
Resources
We’re Engineering the Network with Open Standards for Performance at Every Scale
Footnotes
- Calculations by AMD Performance Labs in June 2025, based on the projected memory capacity/ bandwidth and scale up/out bandwidth specifications of AMD Instinct™ MI455X 72xGPU “Helios” AI Rack vs. the publicly announced NVIDIA “Vera Rubin” 72xGPU “Oberon” Rack. Server manufacturers may vary configurations, yielding different results. MI350-045A
Calculations by AMD Performance Labs in September 2025, based on the FP8/FP4 datatypes and the projected specifications for AMD Instinct™ MI455X 72xGPU “Helios” AI Rack vs. Publicly announced specs for the NVIDIA “Vera Rubin” 72xGPU “Oberon” AI Rack. Actual results based on production silicon may vary. Server manufacturers may vary configurations, yielding different results. MI350-046B
- Based on AMD Engineering silicon modeling and AMD synthetic benchmark simulation, using the MOE_4p5T hi_sparsity_ea9 benchmark test to project time- to- solution speed up in days using the FP8 training datatype on a simulated LLM system modeled with 8000 AMD Instinct MI455X GPUs and (3) versus (2) AMD Pensando Vulcano NICs.
Results reflect analysis of pure training compute (decode layers only) Configuration evaluated with a global batch size of 4096 and a sequence length of 4K, using SwiGLU activation. Results exclude evaluation and checkpointing overhead. FlashAttention v3 with matmul–softmax overlap is assumed. AllReduce and All2All communication costs are fully accounted for and not hidden via tiled compute–communication overlap. Gradient synchronization and FSDP weight prefetch are evaluated across varying levels of overlap to assess scale-out sensitivity. Results assume ideal and not fully optimized real-world behavior and may vary when actual product(s) are released in market. MI400-018
- https://focus.world-exchanges.org/issue/june-2025/market-statistics, AMD Internal Sales Data
Footnotes
- Calculations by AMD Performance Labs in June 2025, based on the projected memory capacity/ bandwidth and scale up/out bandwidth specifications of AMD Instinct™ MI455X 72xGPU “Helios” AI Rack vs. the publicly announced NVIDIA “Vera Rubin” 72xGPU “Oberon” Rack. Server manufacturers may vary configurations, yielding different results. MI350-045A
Calculations by AMD Performance Labs in September 2025, based on the FP8/FP4 datatypes and the projected specifications for AMD Instinct™ MI455X 72xGPU “Helios” AI Rack vs. Publicly announced specs for the NVIDIA “Vera Rubin” 72xGPU “Oberon” AI Rack. Actual results based on production silicon may vary. Server manufacturers may vary configurations, yielding different results. MI350-046B - Based on AMD Engineering silicon modeling and AMD synthetic benchmark simulation, using the MOE_4p5T hi_sparsity_ea9 benchmark test to project time- to- solution speed up in days using the FP8 training datatype on a simulated LLM system modeled with 8000 AMD Instinct MI455X GPUs and (3) versus (2) AMD Pensando Vulcano NICs.
Results reflect analysis of pure training compute (decode layers only) Configuration evaluated with a global batch size of 4096 and a sequence length of 4K, using SwiGLU activation. Results exclude evaluation and checkpointing overhead. FlashAttention v3 with matmul–softmax overlap is assumed. AllReduce and All2All communication costs are fully accounted for and not hidden via tiled compute–communication overlap. Gradient synchronization and FSDP weight prefetch are evaluated across varying levels of overlap to assess scale-out sensitivity. Results assume ideal and not fully optimized real-world behavior and may vary when actual product(s) are released in market. MI400-018 - https://focus.world-exchanges.org/issue/june-2025/market-statistics, AMD Internal Sales Data